CL2 HynSongShin Korea Macropru
-
Upload
angel-munoz -
Category
Documents
-
view
241 -
download
0
Transcript of CL2 HynSongShin Korea Macropru
-
7/28/2019 CL2 HynSongShin Korea Macropru
1/36
Assessing Macroprudential Policies:Case of Korea
Valentina [email protected]
Hyun Song [email protected]
April 12, 2013
Abstract
This paper develops methods for assessing the sensitivity of capital ows to globalnancial conditions, and applies the methods in assessing the impact of macroprudentialpolicies introduced by Korea in 2010. Relative to a comparison group of countries, wend that the sensitivity of capital ows into Korea to global conditions decreased in theperiod following the introduction of macroprudential policies.
JEL codes: F32, F33, F34Keywords: Capital ows, credit booms, macroprudential policy
Prepared for the symposium issue of the Scandinavian Journal of Economics on capital ows. We aregrateful to Maurice Obstfeld and two referees for their comments and guidance on an earlier draft. One of theauthors (Shin) was involved in the design of the macroprudential policy tools introduced in Korea in 2010, andhence is not a disinterested party in the ndings reported in this paper.
1
-
7/28/2019 CL2 HynSongShin Korea Macropru
2/36
1 Introduction
Beginning in June 2010, Korea introduced a series of macroprudential measures aimed at build-
ing resilience against external nancial shocks, especially against its well-known vulnerability
to capital ow reversals in the banking sector and the associated disruptions to domestic nan-
cial conditions. Korea was one of the countries hardest hit in the 1997 Asian nancial crisis,
and was again at the sharp end of the nancial turmoil unleashed after the failure of Lehman
Brothers in September 2008. In recognition of the sources of Koreas vulnerabilities (on which
more below), the macroprudential measures introduced from 2010 were aimed at moderating
the procyclicality of the banking sector by dampening the uctuations in the growth of so-called
non-core bank liabilities, especially cross-border banking sector liabilities.
The purpose of our paper is to give a preliminary empirical assessment of the impact of the
measures introduced by the Korean authorities, and to revisit the rationale behind their design
so as to rene the thinking behind capital ows and nancial stability. In this respect, the case
of Korea represents a natural experiment for investigating the impact of new macroprudential
policies. Our assessment is based on the framework developed in our earlier paper on global
liquidity (Bruno and Shin (2011)) where global nancial conditions drive banking sector capital
ows through the funding and lending operations of international banks. Our empirical proxies
for global nancial conditions draw on the institutional structure of cross-border banking and thestatus of the US dollar as the currency that underpins the global banking system, as explained
below.
Our assessment of the performance of Koreas macroprudential tools is based on a panel study
where Korea is one of 48 countries in a sample that encompasses both advanced and emerging
economies. Our approach is to treat the countries other than Korea as a comparison group and
ask, rst, how Koreas susceptibility to the global factors in capital ows compares to the other
countries during the entire sample period. Then, having obtained a benchmark for comparison
from this cross-country panel study, we ask whether the empirical relationship between Korea
and the comparison group changed in any noticeable way following the sequenced introduction
2
-
7/28/2019 CL2 HynSongShin Korea Macropru
3/36
of macroprudential measures in Korea from June 2010. We exploit the panel structure by
reviewing the evidence both across time and in the cross-section, as well as examining the full
complement of interaction dummies to test for structural changes.
To anticipate our main conclusion, we do indeed nd evidence that capital ows into Korea
became less sensitive to global factors after the introduction of its macroprudential measures.
Interestingly, this change in Koreas sensitivity to global conditions is in contrast to the other
countries in the region. We nd that Koreas experience is the opposite of other comparable
countries in Asia, whose sensitivity to global liquidity conditions actually increased after June
2010. Specically, when we examine the same set of regressions applied one by one to the Big
Five ASEAN countries (Indonesia, Malaysia, Philippines, Thailand and Vietnam) as well as
to Australia, we nd that their incremental sensitivity to global liquidity conditions was higherafter June 2010. Thus, Koreas lower sensitivity to global liquidity conditions after June 2010
stands out in contrast.
More broadly, we see the contribution of our paper as oering a simple but useful methodol-
ogy for examining the impact of macroprudential policies by identifying variables that are known
proxies for global liquidity, and then investigating how the sensitivity of a particular aggregate
to global conditions varies over time, before and after the introduction of the new policies.
Although our empirical methodology cannot fully control for other events in the capital-
recipient economy that occur at the same time as the introduction of the new policy regime,
the use of country xed eects and country-level control variables account for country-specic
shocks. Our method is a useful rst step when searching for instances that deserve more careful
scrutiny through micro empirical investigations.
The outline of the paper is as follows. We begin by describing the background to our study
by outlining the rationale for why the non-core liabilities of the banking sector are a good
proxy for the underlying nancial conditions and the vulnerability to a reversal. We further
explain how, in the context of cross-border banking, capital ows through the banking sector
are closely related to the uctuations in non-core liabilities. We then describe the institutional
background for Korea and outline the timing and sequencing of the macroprudential measures
3
-
7/28/2019 CL2 HynSongShin Korea Macropru
4/36
introduced in Korea. The core empirical investigation of the paper then follows in two sections.
We conclude by drawing implications for the relationship between capital ows and nancial
stability.
2 Background
2.1 Global Banking System
Banking activity is a key driver of nancial conditions both within and across borders. Rapid
growth of bank lending is mirrored on the liabilities side of the balance sheet by shifts in the
composition of bank funding. As banks are intermediaries who borrow in order to lend, they
must seek funding in order to lend to their borrowers. In an economy with domestic savers,the primary source of funding available to the bank is the retail deposits of the household sector
- the core funding. During a credit boom, however, the bank resorts to alternative, non-
core liabilities to nance its lending when its access to core deposit funding does not keep pace
with the growth of its lending.1 Cross-border bank nancing where banks draw on wholesale
funding supplied by the global banks is likely to be an important component of non-core funding
when the nancial system has an open banking sector as in Korea. Given the close connection
between procyclicality and capital ows, there are close parallels between currency crises and
credit crises in countries that operate with open banking sectors. Hahm, Shin and Shin (2011)
nd in their panel probit study of nancial crisis indicators that the ratio of non-core to core
funding (especially the non-core liabilities to foreign creditors) is the most consistently reliable
indicator of vulnerability of a country, both to a currency crisis and to a credit crisis.
Elsewhere (Bruno and Shin (2011)), we have explored the role of global supply push factors
as key determinants of cross-border bank capital ows. In particular, variables that determine
the risk-taking behavior of global banks through their leverage decision turn out to be useful
in explaining the supply push forces in cross-border capital ows. Here, we build on these
earlier ndings by examining how macroprudential policies may mitigate such global supply
1 The distinction between core and non-core funding is discussed in more detail in Shin and Shin (2010).
4
-
7/28/2019 CL2 HynSongShin Korea Macropru
5/36
Regional Bank Global Bank
A A LL
Wholesale
Funding
Market
Local
Borrowers
Stage 1Stage 2Stage 3
Figure 1. Three stages of cross-border banking sector ows.
push forces by moderating the procyclicality of the banking sector.
As a background to our study, we reiterate briey the institutional backdrop that motivates
our approach to cross-border banking as well as provide some additional institutional background
for Korea.2 Our empirical investigation rests on the interaction between local and global banks
depicted in Figure 1. The direction of nancial ows goes from right to left, to stick to the
convention of having assets on the left hand side of the balance sheet and liabilities on the right
hand side. In stage 1 in Figure 1, global banks raise wholesale funding and supply wholesale
funding to local banks in other jurisdictions. The local banks draw on the cross-border funding
(stage 2) in order to lend to their local borrowers (stage 3). Stage 1 corresponds to the activity
of global banks borrowing in nancial centers.A BIS (2010) study describes how the branches and subsidiaries of foreign banks in the United
States borrow from money market funds and then channel the funds to their headquarters.
Baba, McCauley and Ramaswamy (2009), McGuire and von Peter (2009), IMF (2011) and Shin
(2012b) note that in the run-up to the crisis, roughly 50% of the assets of U.S. prime money
market funds were obligations of European banks. The funds channeled by the branch to
headquarters (interoce assets) constitute gross capital outows from the United States.
Figure 5 plots the assets and liabilities of foreign banks in the United States (left panel) and
their net interoce assets (right panel). Net interoce assets measure the net claim of the2 Readers interested in further details may consult Bruno and Shin (2011) and Shin and Shin (2010) for the
background on the global banking system and for Koreas experience, respectively.
5
-
7/28/2019 CL2 HynSongShin Korea Macropru
6/36
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
22-Sep-04
06-Apr-05
19-Oct-05
03-May-06
15-Nov-06
30-May-07
12-Dec-07
25-Jun-08
07-Jan-09
22-Jul-09
03-Feb-10
18-Aug-10
02-Mar-11
14-Sep-11
28-Mar-12
Trillion
dollars
N et in te ro ff ic e a ss et s L arg e t im e d ep os it s B or ro wi ng s f ro m ba nk s i n U .S . B orr owi ng s f ro m ot he rs
Securities Loans and leases Cash assets
-400
-200
0
200
400
600
800
20-Apr-88
17-May-89
13-Jun-90
10-Jul-91
05-Aug-92
01-Sep-93
28-Sep-94
25-Oct-95
20-Nov-96
17-Dec-97
13-Jan-99
09-Feb-00
07-Mar-01
03-Apr-02
30-Apr-03
26-May-04
22-Jun-05
19-Jul-06
15-Aug-07
10-Sep-08
07-Oct-09
03-Nov-10
30-Nov-11
Billion
dollars
Net interoffice assets of foreign banks in US
Figure 2. The left hand chart shows the assets and liabilities of foreign bank branches and subsidiaries (foreign-related institutions) in the US on their parent. The right hand chart shows the net interoce assets of foreign
banks in the US, given by the negative of the net due to foreign-related oces. (Source: Federal Reserve H8series)
branch or subsidiary of the foreign bank on its parent. Normally, net interoce assets would
be negative, as foreign bank branches act as lending outposts. However, we see that the decade
between 2001 to 2011 was exceptional, when net interoce assets turned sharply positive, before
reversing into negative territory during the height of the European crisis in 2011. In eect,
during the decade between 2001 and 2011, foreign bank oces became funding sources for the
parent, rather than lending outposts. As noted by the BIS (2010) report, many Europeanbanks use a centralized funding model in which available funds are deployed globally through a
centralized portfolio allocation decision. Cetorelli and Goldberg (2009, 2010) provide extensive
evidence using bank level data that internal capital markets serve to reallocate funding within
global banking organizations.
The net interoce position of foreign banks in the US therefore reects the extent to which
global banks were engaged in supplying US dollar funding to other parts of the world. In
our empirical investigation below, we will use the growth of the net interoce account position
of foreign banks in the US as a key empirical proxy for the availability of wholesale funding
provided to borrowers in the capital-recipient economy - a key supply push factor.
6
-
7/28/2019 CL2 HynSongShin Korea Macropru
7/36
2008Q1
-20.0
-15.0
-10.0
-5.0
0.0
5.0
10.0
15.0
20.0
1977-Q
4
1979-Q
2
1980-Q
4
1982-Q
2
1983-Q
4
1985-Q
2
1986-Q
4
1988-Q
2
1989-Q
4
1991-Q
2
1992-Q
4
1994-Q
2
1995-Q
4
1997-Q
2
1998-Q
4
2000-Q
2
2001-Q
4
2003-Q
2
2004-Q
4
2006-Q
2
2007-Q
4
2009-Q
2
2010-Q
4
2012-Q
2
Trillion
dollars
Assets
Liabilities
Other
Swiss Franc
Yen
Sterling
Euro
US dollar
Figure 3. Foreign currency assets and liabilities of BIS reporting banks, classied according to currency (Source:BIS Locational Banking Statistics Table 5A)
The reason for our focus on US dollar-denominated bank ows stems from the dominant
role played by the US dollar in the global banking system. Figure 3 plots the foreign currency
assets and liabilities of banks globally, as measured by the BIS locational banking statistics.
Locational data are organized according to the residence principle, and so the US dollar seriesin Figure 3 show the US dollar-denominated assets and liabilities of banks outside the United
States. The Euro series show the corresponding Euro-denominated assets and liabilities of
banks that are outside the Euro area, and so on.
What is clear from Figure 3 is that the US dollar is the dominant currency for international
banking, and has been the currency behind the growth of gross capital ows highlighted by
Borio and Disyatat (2011) and Obstfeld (2012). The US dollar asset series exceeded 10 trillion
dollars in 2008Q1, briey exceeding the total assets of the US chartered commercial bank sector
(see Shin (2012b)). The role of the other currencies are much smaller in comparison. For thisreason, we may expect the net interoce assets of foreign banks in the United States to be a
7
-
7/28/2019 CL2 HynSongShin Korea Macropru
8/36
Dec 2008
Mar 2003 =100
0
50
100
150
200
250
300
350
400
450
500
Mar.1999
Dec.1999
Sep.2000
Jun.2001
Mar.2002
Dec.2002
Sep.2003
Jun.2004
Mar.2005
Dec.2005
Sep.2006
Jun.2007
Mar.2008
Dec.2008
Sep.2009
Jun.2010
Mar.2011
Dec.2011
Ireland
Spain
Turkey
Australia
South Korea
Chile
Brazil
South Africa
Figure 4. External claims (loans and deposits) of BIS reporting country banks on borrowers in countries listed.The series are normalized to 100 in March 2003 (Source: BIS Locational Banking Statistics, Table 7A)
key indicator of the availability of US dollar funding for cross-border transactions, and we pay
special attention to this variable in our empirical investigation.
Stage 2 in Figure 1 corresponds to the cross-border capital ows through the banking sector.
The empirical counterpart of Stage 2 in Figure 1 in our paper will be the cross-border claims
of the banks in countries that report loan amounts to the Bank for International Settlements
(BIS). Figure 4 plots the cross-border claims of BIS-reporting banks on counterparties listed
in the countries on the right. The series have been normalized to equal 100 in March 2003.
Although the borrowers have wide geographical spread, we see a synchronized boom in cross-
border lending before the recent nancial crisis.
The observed capital ows reect the interaction of the supply and demand for wholesale
funding between global and local banks. When local and global banks interact in the market
for wholesale bank funding, the liabilities of local banks serve as the assets of the global banks,
and the lending by global banks is the supply of wholesale funding, while the borrowing by local
banks is its demand. The distinction between the demand and supply of wholesale funding
harks back to Calvo, Leiderman and Reinhart (1996), who distinguished the push and pull
8
-
7/28/2019 CL2 HynSongShin Korea Macropru
9/36
factors that drive capital ows into emerging economies. Nevertheless, to the extent that global
supply push factors are important determinants of capital ows that aect all capital recipient
countries, macroprudential policies that mitigate the procyclical response of an economy to such
factors may have stabilization benets. The fact that global factors aect all countries also
allows us to exploit the panel structure of the data and examine how the sensitivity of capital
ows into a particular country varies before and after the introduction of new policies by treating
the other countries in the sample as a comparison group.
In our empirical investigation, we make use of two global factors. The rst is the growth of
the net interoce assets of foreign banks in the United States (the series shown in the right hand
panel of Figure 5), reecting the activities of international banks that engage in the supply of
wholesale bank funding. A rapid increase in the net interoce assets series reects an expansionof cross-border banking activities of global banks. We will see that the run-up in cross-border
lending in Figure 4 closely mirrors the increase in wholesale funding raised by the global banks
in Figure 5 In eect, Figure 5 reects the liabilities side of global banks balance sheets (Stage
1 in Figure 1), while Figure 4 traces the movements on the asset side of global banks balance
sheets (Stage 2 in Figure 1).
The second set of global factors we employ in our empirical investigation are those associated
with the VIX index of implied volatility of equity index options in the United States. There is
well-documented evidence that banking sector leverage is closely associated with uctuations in
the VIX index (see, for instance, Adrian and Shin (2010, 2012)). Leverage of the banking sector
- both global and local - are important determinants of cross-border claims. When leverage is
high, an additional unit of bank capital will translate into a higher level of cross-border claims.
In addition, any increase in bank leverage will mean that existing bank capital will support
higher amounts of lending. Therefore, since VIX is correlated with bank leverage, the theory
predicts that both the level of the VIX, as well as the change in the VIX will show up as being
determinants as capital ows (see Bruno and Shin (2011)).
9
-
7/28/2019 CL2 HynSongShin Korea Macropru
10/36
2005/0
1
2005/0
5
2005/0
9
2006/0
1
2006/0
5
2006/0
9
2007/0
1
2007/0
5
2007/0
9
2008/0
1
2008/0
5
2008/0
9
2009/0
1
2009/0
5
2009/0
9
2010/0
1
2010/0
5
2010/0
9
2011/0
1
2011/0
5
2011/0
9
2012/0
1
2012/0
5
2012/0
9
2013/0
1
-22.0
-17.0
-12.0
-7.0
-2.0
3.0
8.0
13.0 Billion
Dollars
Bank long-term loan liabilities Bank long-term debt securitiesBank short-term loan liabilities Bank short-term debt securitiesCentral bank deposit liabilities
June 2010
Lehman
Crisis
Figure 5. Capital ows to the Korean banking sector, Jan 2005 - Feb 2013 (monthly) (Source: Bank of Korea)
2.2 Case of Korea
The procyclicality of the banking sector and its use of cross-border funding is a useful lens
through which to view Koreas experience. Korea was one of the countries hardest hit during
the 1997 Asian nancial crisis, and its experience leading up to the 2008 nancial crisis isrevealing in several respects. From 2005 to 2007, locally-owned Korean banks and the foreign
bank branches in Korea saw rapid increases in short-term foreign currency liabilities, which then
subsequently reversed abruptly after the bankruptcy of Lehman Brothers in September 2008.
The severity of the 2008 crisis for Korea can be attributed largely to the rapid deleveraging
that took place by the banking sector (both domestic and foreign) with the onset of the 2008
nancial crisis. Figure 5 plots the capital ows to the Korean banking sector from the Bank of
Koreas monthly balance of payments statistics.
We see that the Korean banking sector saw rapid inows in the form of short-term loan
liabilities which turned into substantial outows once the crisis hit in September 2008, associated
10
-
7/28/2019 CL2 HynSongShin Korea Macropru
11/36
with the kind of deleveraging of the banking sector that has become the hallmark of the global
nancial crisis in 2008. Deleveraging sets o amplifying eects through price changes, where
the depreciation of the Korea won against the US dollar increases the size of external liabilities
relative to domestic currency assets. In addition, the creditor banks in global capital markets
are similarly shrinking their lending in response to heightened measures of risk. The feedback
loop generated by such reactions to price changes can lead to amplication of shocks.3 In the
three months following the Lehman bankruptcy, the outow of short-term liabilities from the
banking sector was in excess of 50 billion dollars and largely accounts for the net decrease in
Koreas foreign exchange reserves from over 240 billion dollars before the Lehman crisis to 200
billion at the end of 2008. We also see in Figure 5 the operations of the Bank of Korea, which
entered into a 30 billion dollar swap agreement with the Federal Reserve in October, which wassubsequently unwound following the acute stage of the crisis.
During the period of rapid capital inows, the banking sector in Korea (including the foreign
bank branches) also held dollar assets, but the counterparties were local borrowers, such as
exporting companies who held long-term dollar assets arising from their export receivables.
Although the overall currency mismatch on the consolidated balance sheet consisting of the
corporate and banking sectors would then cancel out, a maturity mismatch between long-term
dollar claims and short-term dollar liabilities took its place. In eect, the currency mismatch
was replaced by a maturity mismatch which left the Korean nancial system vulnerable to the
global nancial crisis in 2008 that followed in the wake of the bankruptcy of Lehman Brothers.
Chung, Park and Shin (2012) give more detailed discussion of the role of hedging by exporting
companies as a contributory factor in the rapid growth of short-term foreign currency bank
liabilities.
These lessons led to a concerted policy initiative on the part of Korean policy makers to
mitigate some of the known vulnerabilities. The IMF Background Paper on macroprudential
3 Theoretical developments of such feedback eects through prices and heightened risks can be found in Xiong(2001), Brunnermeier and Pedersen (2009), Danielsson, Shin and Zigrand (2011) and Brunnermeier and Sannikov(2011).
11
-
7/28/2019 CL2 HynSongShin Korea Macropru
12/36
policies (IMF (2012)) provides information on the timing and rationale of the macroprudential
policies in Korea. Beginning in June 2010, the authorities in Korea introduced a sequence of
macroprudential measures aimed at building resilience against its well-known vulnerability to
capital ow reversals in the banking sector and the associated disruptions to domestic nancial
conditions. The policy initiative was widely reported in both the domestic and international
press at the time, and the press reports are useful in dating the sequence of events.4
The rst policy measure announced by the Korean authorities (in June 2010) was a leverage
cap on the notional value of foreign exchange derivatives contracts (encompassing currency swaps
and forwards) that banks could maintain (see IMF (2012, p. 50)). For foreign bank branches,
the leverage cap was set at 2.5 times their capital, while for domestic Korean banks, the cap
was set at 50% of their capital. Foreign banks could in principle increase their positions byallocating greater capital to their branches in Korea, but the leverage cap lowers the return to
capital for banks engaged in this segment of their business, thereby serving as a disincentive on
expansion of derivatives positions.
The second component was the levy on the non-core liabilities of the banks (the macro-
prudential levy), applied to the foreign exchange-denominated liabilities of the banking sector.
The Korean non-core liabilities levy was relatively unfamiliar compared to the standard bank
capital-related tools or standard capital control tools such as the unremunerated reserve re-
quirements (URR). For this reason, the roll-out took more time. Although the policy was
discussed from February of 2010 (Shin (2010)) and press coverage trailed the introduction of
the non-core levy from early in 2010,5 the measure was announced formally in December 2010,
after the conclusion of the G20 Seoul summit in November. The legislation was passed in April
of 2011, and the levy began its operation in August 2011 (see IMF (2012)).
The levy consists of an annualized 20 basis point charge on the wholesale foreign exchange
denominated liabilities of the banks of maturity up to 12 months. Lower rates are applied in a
4 See, for instance, the June 2010 article in the Economist magazine, The won that got away: a surgicalstrike in a volatile market http://www.economist.com/node/16381310
5 See Wall Street Journal, April 22, 2010:http://blogs.wsj.com/economics/2010/04/22/is-obamas-bank-tax-plan-right-for-emerging-markets/
12
-
7/28/2019 CL2 HynSongShin Korea Macropru
13/36
graduated manner to maturities of over one year.
The levy was designed so that the proceeds of the levy are paid into a special segregated
account of the foreign exchange reserves, rather than going into the general revenue of the
government. In this respect, the Korean levy was designed from the outset as a nancial
stability tool, rather than as a scal measure. The outwardly similar bank levies introduced
by France and the UK in 2010 had the proceeds being paid into general government revenue,
and were designed as scal measures to supplement government revenue. By targeting non-core
liabilities only, the levy was also designed to address the procyclicality of the banking sector
while leaving unaected (as far as possible) the intermediation of core funding from savers to
borrowers.
Figure 5 indicates the date of the introduction of the macroprudential policies by the darkgrey bar at June 2010. We see some evidence from the chart that short-term bank liabilities
continued to shrink, and was replaced with long-term liabilities, both in the form of long-term
securities and long-term loans. However, just examining the series for Korea in isolation does
not control for external conditions and other factors that aect the banking landscape more
broadly. For a proper assessment of the policies, we need to examine Koreas experience in
comparison with other countries. This is our task in the rest of the paper.
For the purpose of our empirical investigation, the exact dating of the impact of the non-
core liabilities levy is complicated by the long gestation period between initial discussions of its
adoption and its nal implementation. Some anticipation of the new measure will have aected
behavior before the formal introduction of the levy, but the quantication of the anticipation
eect is dicult, and is not attempted here. The announcement of the cap of foreign exchange
derivatives in June 2010 serves as a useful threshold point, but we will conduct a number of
tests to enable us to identify the break point in the time series.
The Korean measures should be seen in the context of the broader debate on macroprudential
policies. Galati and Moessner (2012) survey the recent literature, and Claessens and Ghosh
(2012) discuss evidence on the eectiveness of macroprudential policies for moderating credit
growth in emerging economies. As a policy tool, the Korean non-core liabilities measure is
13
-
7/28/2019 CL2 HynSongShin Korea Macropru
14/36
distinctive in the way that it acts directly on cost of bank liabilities, unlike bank capital-related
tools or direct restraints on credit growth, such as loan-to-value (LTV) or debt service-to-
income (DTI) caps. Shin (2012a, section 4) discusses the advantages of such an approach and
the broader tradeos involved in alternative macroprudential policies. We do not pursue the
policy design implications here, but our empirical results may be useful in more detailed and
comprehensive studies of the tradeos involved in macroprudential policy design.
3 Empirical Findings
3.1 Data Description
In conducting our assessment of the impact of Koreas macroprudential policies, our sample ofcountries includes all the developed countries, the "Big Five" countries in ASEAN (so as to
provide the basis for a regional comparison for Korea) and a selection of emerging economies.
The selection of developing countries is based on the Claessens, van Horen, Gurcanlar and
Mercado (2008) database on foreign banks. We select countries where foreign banks play an
economically signicant role in the countrys nancial system but we exclude oshore nancial
centers.
The full list of countries included in our sample are Argentina, Australia, Austria, Belgium,
Brazil, Bulgaria, Canada, Chile, Cyprus, Czech Republic, Denmark, Egypt, Estonia, Finland,
France, Germany, Greece, Hungary, Iceland, Indonesia, Ireland, Israel, Italy, Japan, Latvia,
Lithuania, Malaysia, Malta, Mexico, Netherlands, Norway, Philippines, Poland, Portugal, Ro-
mania, Russia, Slovakia, Slovenia, South Korea, Spain, Sweden, Switzerland, Thailand, Turkey,
Ukraine, United Kingdom, Uruguay and Vietnam. Table 1 gives the main summary statistics
of our sample of 48 countries.
Our measure of capital ows is the quarterly growth in external claims of BIS reporting
country banks, as given by the BIS locational statistics (Table 7A). As explained above, the
growth in the interoce assets of foreign banks in the United States (from the Federal Reserve,
series H8 on commercial banks) and the VIX (the CBOE index of implied volatility in S&P
14
-
7/28/2019 CL2 HynSongShin Korea Macropru
15/36
Table 1. Summary Statistics. This table summarizes our key variables in terms of their, mean, standarddeviation, minimum and maximum.
Obs Mean Std. Dev Min Max
Interoce 65 0.072365 0.347339 -0.67924 0.785736VIX 65 3.040451 0.346168 2.432736 3.787366
VIX 65 0.003285 0.297389 -0.60753 0.955698RER 2812 -0.00261 0.062672 -0.25199 1.03038
Money stock 65 0.0137911 0.0294825 -0.047228 0.0863669GDP Growth 691 0.075439 0.068993 -0.20757 0.415204
Debt to GDP 691 -0.0016 0.100881 -0.17825 0.220988BIS Loans growth 2812 0.025885 0.095853 -0.25857 0.340927
500 stock index option prices) are the two proxies reecting the cross-border activities and the
leverage of global banks. In addition to the VIX in log level, we also include the quarterly log
dierence of the VIX (denoted by VIX). The VIX level is a proxy for the leverage of the global
banks, and hence the rate at which each unit of new equity is translated into lending. On the
other hand, VIX is the proxy for the change in leverage, and hence the rate at which lending
grows based on the existing level of equity (see Bruno and Shin (2011) for details).
We also include several control variables - both global and local - as possible determinants of
capital ows. We include the log real exchange rate (RER), where RER is computed as the log
of nominal exchange rate*(US CPI/local CPI). The nominal exchange rate is in units of nationalcurrency per U.S. Dollar (from the IMFs IFS database). The quarterly growth in the global
money supply (Money stock) is calculated as the quarterly log dierence of the sum of the M2
stock in the US, Eurozone and Japan and M4 in the UK (from the IFS). GDP growth is the
country percentage change in GDP from the previous year (from the WEO). Debt to GDP
is the change in government gross debt to GDP (from WEO). All quarter variables (with the
exception ofVIX) are lagged by one quarter to mitigate endogeneity issues. The sample period
spans from the rst quarter of 1996 (the rst date covered in Table 7A of the BIS locational
data) to the latest data at the time of writing, which is the rst quarter of 2012.
15
-
7/28/2019 CL2 HynSongShin Korea Macropru
16/36
3.2 Empirical Specication
Our empirical investigation is based on various modications of the following benchmark panel
regressions similar to Bruno and Shin (2011):
Lc;t = + 1 Interocet1 + 1VIXt1
+1VIXt + controls + ec;t (1)
Here, Lc;t is banking sector capital inow into country c in period t, as given by the quarterly
log dierence in the external claims of BIS reporting country banks on country c between quarters
t and t 1; VIXt1 is the log of end-quarter VIX index lagged by one quarter; Interocet1 is
the growth in interoce assets of foreign banks in the US from the quarter before given by the
percentage growth and lagged by one quarter. The control set includes all the variables listedin Section 3.1. Regressions include country xed eects, year dummies and clustered standard
errors at the country level. Year dummies still leave scope for quarterly variations in our global
variables, and they turn out to be highly signicant, as seen below.
Our empirical approach is to include in the benchmark panel specication a dummy variable
equal to 1 (0 otherwise) for the period from June 2010 (Post 2010) and a dummy variable
equal to 1 (0 otherwise) for Korea (Korea). We then interact each dummy variable with each
global factors. Our focus is on the triple interaction terms given by
Global factor Korea Post 2010 (2)
which give the incremental sensitivity of capital ows to Korea to the global factor from June
2010. In addition, we also examine the coecients of the pairwise interaction terms:
Global factor Korea, Global factor Post 2010 (3)
By comparing the coecients between the triple interaction term (2) and the two double inter-
action terms (3), we can ascertain whether the change in sensitivity to the global factor in Korea
after 2010 is due to a shift over time, or whether the change in sensitivity is Korea-specic. Wealso construct similar interaction terms and regressions for all the other variables included in
the benchmark estimation.
16
-
7/28/2019 CL2 HynSongShin Korea Macropru
17/36
4 Empirical Findings
4.1 Sensitivity to Global Financial Activities
Table 2 presents our rst set of panel regressions. The structure of the interactions can be
usefully summarized in the following 2 2 matrix for the case of the Interoce variable,
which gives the total eects resulting from the impact of the Interoce variable for Korea and
non-Korean countries for the periods before and after June 2010.
Korea 1 + 3 1 + 2 + 3 + 4
Others 1
1
+ 2
Pre 2010 Post 2010
(4)
The coecient 1 in Table 2 measures the impact ofInteroce before June 2010 on capital
ows for all countries, except Korea, while 2
measures the incremental impact of Interoce
after June 2010. The coecient 3 measures the incremental impact of Interoce on Korea
before June 2010 , while 3 + 4 is for the period after June 2010. There is an analogous set of
interactions terms associated with the other two global factors - VIX ( coecients) and VIX
( coecients). In Table 2, column (3) gives the i estimates for the interaction terms with
the Interoce variable, and column (5) gives the and coecients for the interaction terms
involving the VIX and VIX global factors.
Column (1) reports results from the benchmark regression (1). The VIX index in levels and
log dierences as well as the Interoce variables are highly signicant and of the predicted
sign as also shown in Bruno and Shin (2011). The signicance of Interoce indicates that
the activities of global banks (mainly European banks, as shown in Shin (2012b)) are correlated
with the capital inows into our sample of countries. In this context, uctuations in the VIX
index (both in the level as well as its quarterly log dierence) are associated with shifts in theleverage of the banking sector and hence the capital ows through the banking sector.
17
-
7/28/2019 CL2 HynSongShin Korea Macropru
18/36
Table 2. Panel regressions on capital ows. This table presents panel regressions for bank capital ows to48 countries. The dependent variable is bank capital ows measured by the quarterly log dierence of externalloans (BIS Table 7A). Explanatory variables include the growth in interoce assets, the VIX and the changein the VIX and their interactions with a time dummy for the period from June 2010 and a dummy for Korea.See text for explanation of methodology. Standard errors are clustered at the country level and are reported inparantheses. Data are for 1996Q1-2012Q1.
Coe. 1 2 3 4 51 Interoce 0.0152*** 0.0148*** 0.0103* 0.0151*** 0.0130**
[0.0053] [0.0054] [0.0053] [0.0053] [0.0053]2 Interoce*Post 2010 0.0569**
[0.0222]3 Interoce*Korea 0.0164*** 0.0360***
[0.0048] [0.0053]4
interoce*Korea*Post 2010 -0.1160***[0.0160]
1
VIX -0.0502*** -0.0503*** -0.0451*** -0.0490*** -0.0607***
[0.0088] [0.0088] [0.0089] [0.0088] [0.0098]2 VIX*Post 2010 0.0139***
[0.0038]3
VIX*Korea -0.0674*** -0.0679***[0.0060] [0.0058]
4
VIX*Korea*Post 2010 -0.0082[0.0098]
1 VIX -0.0217*** -0.0217*** -0.0171** -0.0217*** -0.0160*[0.0077] [0.0077] [0.0079] [0.0078] [0.0087]
2 VIX*Post 2010 -0.0235**[0.0107]
3 VIX*Korea -0.0063 -0.0231***[0.0059] [0.0065]
4 VIX*Korea*Post 2010 0.0378***[0.0104]
Korea*Post 2010 -0.0131 0.0321[0.0095] [0.0331]
RER -0.0765** -0.0760** -0.0791** -0.0741** -0.0736**[0.0299] [0.0298] [0.0296] [0.0294] [0.0289]
Money stock -0.0413 -0.0408 -0.0629 -0.0387 -0.0481[0.0659] [0.0659] [0.0660] [0.0661] [0.0665]
GDP Growth 0.2230*** 0.2227*** 0.2230*** 0.2250*** 0.2248***[0.0821] [0.0822] [0.0821] [0.0828] [0.0829]
Debt to GDP -0.0667** -0.0669** -0.0661** -0.0686** -0.0686**[0.0311] [0.0311] [0.0313] [0.0315] [0.0318]
Constant 0.1493*** 0.1495*** 0.1439*** 0.1502*** 0.1356***
[0.0273] [0.0273] [0.0270] [0.0273] [0.0269]Observations 2,812 2,812 2,812 2,812 2,812R-squared 0.127 0.127 0.13 0.129 0.134
Number of countries 48 48 48 48 48Time and country xed eects Y Y Y Y Y
18
-
7/28/2019 CL2 HynSongShin Korea Macropru
19/36
Columns (2) and (3) shows results of our augmented specication with the dummy Korea and
the dummy Post 2010 and their interactions with Interoce. We see that 2 > 0, indicating
that the impact of the global factor Interoce is stronger after June 2010 than before. Also,
we have 3> 0 indicating that Korea has traditionally been more sensitive to Interoce than
other countries. However, we see that 4 < 0, meaning that Koreas sensitivity to Interoce
has fallen after June 2010, in spite of Koreas sensitivity in the past and in spite of the greater
impact of Interoce for other countries.
Columns (4) and (5) show results of our augmented specication with the dummy Korea
and the dummy Post 2010 and their interactions with VIX and VIX. Column (5) shows that
Korea has had greater sensitivity to the other global factors, VIX and VIX, as seen from the
fact that 3 < 0 and 3 < 0 (recall that VIX and leverage are inversely related, and so morenegative coecients indicate greater sensitivity). Nevertheless, 4 is not signicant and 4 < 0
indicating that Korea hasnt become more sensitive to the VIX and actually has become less
sensitive to VIX and VIX after June 2010.
We also conduct a full set of F-tests on the incremental sensitivity and total eect over
time and across countries. From the 2 2 table in (4), the columns indicate the cross-section
dimension of Korea versus the rest of the sample, while the rows correspond to the time series
dimension of pre- and post-2010. The key tests highlighting the dierence in the Interoce
eect pre and post 2010 for Korea and the dierence pre and post 2010 between Korea and the
rest of the countries are the following:
2 + 4 = 0 tests the null hypothesis that there has been no change in the sensitivity of
Korea to Interoce before and after 2010 (comparing across rows in (4)).
3 + 4 = 0 tests the null hypothesis that there is no dierence between Korea and the
rest of the sample to Interoce after 2010 (comparing across columns in (4)).
We perform a similar set of tests also for VIX (coecients ) and VIX (coecients ):Table
3 reports the full set of F-tests for Interoce, VIX and VIX each corresponding to the
19
-
7/28/2019 CL2 HynSongShin Korea Macropru
20/36
Table 3. F tests for coecient restrictions in Table 2. This table presents F tests for coecient restrictionsin Table 2. The column labels refer to Table 2 and p-values are reported for each null hypothesis. The null2
+ 4
= 0 is the statement that there is no change in sensitivity of capital ows into Korea with respect toInteroce after June 2010. The null
3+
4= 0 corresponds to the statement that there is no dierence
between Korea and other countries in their sensitivity to Interoce after June 2010. and coecients arefor the other global factors VIX and VIX, respectively.
Null hypothesis
(p-values reported)1 2 3 4 5
2
+ 4
= 0 0.00003 + 4 = 0 0.00001 + 2 = 0 0.00431 + 3 = 0 0.0000 0.0000
1 + 2 + 3 + 4 = 0 0.3162
2
+ 4
= 0 0.58923 + 4 = 0 0.00001 + 2 = 0 0.0000
1 + 3 = 0 0.0000 0.0000
1 + 2 + 3 + 4 = 0 0.00002 + 4 = 0 0.03873 + 4 = 0 0.11511 + 2 = 0 0.00051 + 3 = 0 0.0000 0.0000
1 + 2 + 3 + 4 = 0 0.8098
20
-
7/28/2019 CL2 HynSongShin Korea Macropru
21/36
respective columns-specications in 2. For instance, the tests of the null hypothesis that there
is no dierence for Koreas sensitivity to the global banking activities after 2010 from the rest
of the sample are 3 + 4 = 0, 3 + 4 = 0, and 3 + 4 = 0.
While non-Korean countries became more sensitive to changes in Interoce after June 2010,
Table 3 column (3) shows that Korea became less sensitive (2 + 4 negative and signicant).
Moreover, before 2010 the impact of Interoce was higher for Korea than for the rest of the
world (3 positive and signicant) and became lower after 2010 (3+4 negative and signicant).
The total resulting eect for Korea after 2010 is that global banks activities did no longer have
a signicant impact on capital ows for Korea (1 + 2 + 3 + 4 = 0).
A similar trend occurs for the change in VIX. Korea became less sensitive to changes in the
VIX after 2010 (2 + 4 positive and signicant ), whereas non-Korean countries became moresensitive (2 negative and signicant). The total eect for Korea after 2010 is that changes
in VIX did no longer have a signicant impact on capital ows for Korea (1 + 2 + 3 + 4
= 0). However, after June 2010 Korea continued remaining more sensitive to the level of the
VIX than the rest of the world (3 + 4 negative and signicant). Taken together, the weight
of the evidence points to a structural shift in Koreas sensitivity to global banking ows, with
the eect being stronger through the lens of the funding activities (interoce assets) of global
banks.
4.2 Sensitivity to Additional Factors
So far we have examined the impact of global banking activities and how they inuence capital
ows. We now turn to the additional explanatory variables listed in Table 2. In particular, we
see that RER coecient is negative and highly signicant. Hence, when the local currency
appreciates against the dollar from quarter t 1 to quarter t, there is an acceleration of capital
ows into that country from quarter t to quarter t+ 1. This feature is a natural consequence of
a setting where borrowers from the local banks have a currency mismatch, as shown in Bruno
and Shin (2011). When the local currency appreciates, the borrowers balance sheets become
stronger, thereby reducing the credit risk on the banks loan book, which in turn increases the
21
-
7/28/2019 CL2 HynSongShin Korea Macropru
22/36
capacity of local banks to lend. The increased lending is nanced with capital inows.
Capital ows are also increasing in the growth of the global M2 money stock worldwide,
which reects the greater demand for funding by banks, as well as the increased capital ows
through non-nancial rms, who hold the proceeds of any increased funding operation by holding
cash in the banks.
In Table 4, we conduct an analysis of the structural change in the capital ows with respect
to the RER, M2, GDP growth and change in DEBT/GDP similar to that in Table 2. We
follow the analogous method of dening triple interactions between each local factor with the
time dummy that takes the value 1 if the date is June 2010 or later, and with the country
dummy that takes the value of 1 for Korea and zero otherwise. The F-tests for structural
change with respect to each of the four variables are presented in Table 5 , where (1), (2), (3),(4) corresponds to the coecients listed in 4 for each of the four variables RER, M2, GDP
growth and DEBT/GDP growth.
In particular, testing the null hypothesis (3) + (4) = 0 corresponds to the testing the dierent
behavior of Korea with respect to the other countries after June 2010. This null is not rejected
for RER, suggesting little evidence that the relationship between currency appreciation and
capital inows is dierent for Korea relative to other countries after 2010. However, since
capital ow sensitivity has moderated in Korea after 2010 (from our earlier ndings), a possible
explanation is that exchange rate volatility itself has been subdued since the introduction of
the macroprudential measures in 2010. In this sense, the failure to reject the null (3) + (4) =
0 may suggest additional benets to the macroprudential policies that work through exchange
rate stabilization.
In Korea, capital ows increases became less dependent on increases in global money stock
or GDP growth and on decreases in government debt. In particular, Korea was more sensitive
to global money (Money stock) before June 2010 and becomes less sensitive after June 2010,
whereas the eect for the rest of the world is unchanged. An increase in GDP was a signicant
local pull factor before June 2010 for all countries and even more for Korea, but it is actually
reversed after June 2010. Symmetrically, an increase in Debt to GDP was a signicant push
22
-
7/28/2019 CL2 HynSongShin Korea Macropru
23/36
Table 4. Panel regressions with interaction dummies for local variables. This table summarizes panelregressions for bank capital ows to 48 countries. The dependent variable is bank capital ows measured bythe quarterly log dierence of external loans (BIS Table 7A). Explanatory variables include local demandpull factors, M2 growth and their interactions with a time dummy for the period from June 2010 and a dummyfor Korea. Standard errors are clustered at the country level and are reported in parantheses. Data are for1996Q1-2012Q1.
Coe. 1 2 3 4Interoce 0.0139*** 0.0154*** 0.0146*** 0.0154***
[0.0052] [0.0054] [0.0054] [0.0054]VIX -0.0610*** -0.0530*** -0.0534*** -0.0492***
[0.0104] [0.0091] [0.0093] [0.0089]VIX -0.0259*** -0.0244*** -0.0238*** -0.0210***
[0.0079] [0.0079] [0.0081] [0.0077]1 RER -0.0890*** -0.0808*** -0.0707** -0.0786**
[0.0325] [0.0300] [0.0300] [0.0303]2 RER*Post 2010 0.2398***
[0.0830]3
RER*Korea -0.1917***
[0.0334]4 RER*Korea*Post 2010 0.1072
[0.0681]1 Money stock -0.0184 -0.0304 -0.0339 -0.044
[0.0687] [0.0741] [0.0672] [0.0660]2 Money stock*Post 2010 -0.2803
[0.2175]3 Money stock*Korea 0.6394***
[0.0586]4 Money stock*Korea*Post 2010 -0.9889***
[0.1949]1 GDP Growth 0.2240*** 0.2231*** 0.1993** 0.2219**
[0.0822] [0.0821] [0.0852] [0.0840]2 GDP Growth*Post 2010 0.2800**
[0.1119]
3 GDP Growth*Korea 0.4477***[0.0824]
4 GDP Growth*Korea*Post 2010 -2.3514***[0.1996]
1 Debt to GDP -0.0662** -0.0658** -0.0615** -0.0501[0.0317] [0.0313] [0.0299] [0.0326]
2 Debt to GDP*Post 2010 -0.1111[0.0862]
3 Debt to GDP*Korea -0.1579***[0.0328]
4 Debt to GDP*Korea*Post 2010 1.6839***[0.2073]
Korea*Post 2010 0.005 0.0170** 0.1668*** -0.0121[0.0086] [0.0080] [0.0151] [0.0077]
Constant 0.1772*** 0.1542*** 0.1440*** 0.1493***
[0.0318] [0.0277] [0.0270] [0.0276]Observations 2,812 2,812 2,812 2,812R2 0.131 0.129 0.132 0.129
Number of countries 48 48 48 48Year and country xed eects Y Y Y Y
23
-
7/28/2019 CL2 HynSongShin Korea Macropru
24/36
Table 5. F tests for coecient restrictions in Table 4. This table presents F tests for coecient restrictionsin Table 4 for the local variables RER, M2, GDP growth and Debt/GDP. The null (2) + (4) = 0 correspondsto the statement that the sensitivity of capital ows into Korea with respect to the local variable does not changeafter June 2010. The null hypothesis (3) + (4) = 0 is the statement that there is no dierence in sensitivityto that variable between Korea and the other countries after June 2010. p-values are reported for each nullhypothesis.
(1) (2) (3) (4)
Null hypothesis Variable
RER M2 GDP Growth Debt/GDP(2) + (4) = 0 0.0000 0.0000 0.0000 0.0000
(3) + (4) = 0 0.1876 0.0588 0.0000 0.0000
(1) + (2) = 0 0.0624 0.1190 0.0001 0.0637
(1) + (3) = 0 0.0000 0.0000 0.0000 0.0000
(1)+(2)+(3)+(4)=0 0.1855 0.0000 0.0000 0.0000
factor for Korea before June 2010 and it has a reversed eect aect after June 2010, with the
other countries unaected. Overall, these tests strongly show that Korea has indeed a dierent
sensitivity to global and local variables than the rest of the world after June 2010.
4.3 Alternative Test of the Sensitivity to Capital Flows
An alternative approach to testing for the moderation of sensitivity of capital ows is to investi-
gate the reduced co-movement between capital ows into Korea and the aggregate capital owsto all countries. We implement this test with the following two step procedure.6
In the rst step, we compute the variable Beta (i; t) by regressing the capital ows into
country i on the aggregate capital ows to all countries on a rolling eight quarter window
ft 7; ; t 1; tg. The estimation window of Beta (i; t) starts in 1999 and it excludes the
period of the Asian crisis 1997 and 1998 to mitigate endogeneity problems.
In the second step, we use Beta (i; t) as the dependent variable in an OLS regression. The
right hand side variables are the Korea dummy, the Post 2010 dummy and the interaction term
Korea*Post 2010. Other explanatory variables are VIX, VIX. We include the full set of year
6 We are grateful to a referee for suggesting this procedure.
24
-
7/28/2019 CL2 HynSongShin Korea Macropru
25/36
Table 6. Covariation between individual country and aggregate capital ows. This table reportsresults of the two step procedure where rst, we compute Beta (i; t) by regressing capital ows to country i onaggregate capital ows over a rolling eight quarter window ft 7; ; t 1; tg. In the second step, Beta (i; t)is the dependent variable in an OLS regression with global explanatory variables and dummies for Korea andPost-2010 period and their interaction. A full set of time and country dummies are also included. Robuststandard errors clustered at the country level are reported in parantheses.
Dependent variable: Beta(i; t)1 Korea 0.2681***
[0.0406]
2
Korea*Post 2010 -0.7295***
[0.1727]
Post 2010 0.0606
[0.1213]
VIX 0.1623*
[0.0860]
VIX 0.0549[0.0606]
Constant 0.4132
[0.2767]
Observations 1,632
R2 0.308Number of countries 48
Time and country dummies Y
1 + 2 = 0 (p-value) 0.0011
and country dummies, and compute robust standard errors clustered at the country level.
The results are presented in Table 6. The coecient 1 of the Korea dummy indicates the
additional sensitivity of capital ows into Korea relative to the other countries in the sample
before June 2010. We see from Table 6 that the coecient 1 is positive and signicant.
However, we see from the negative coecient 2
on the interaction dummy Korea*Post 2010
that the sensitivity of capital ows into Korea dropped sharply after June 2010.
Even more dramatically, we have 1+2 < 0, so that the sensitivity of Korea ips from being
more sensitive relative to the whole sample to less sensitive. The null hypothesis 1 + 2 = 0
can be rejected with a p-value of 0.0011. This nding casts the macroprudential policies into
an even more favorable light.
25
-
7/28/2019 CL2 HynSongShin Korea Macropru
26/36
Overall, the conjunction of the ndings from the analysis of sensitivity to the global variables
and the local variables point to the moderation of capital ows. We interpret our ndings as
providing some preliminary evidence that the newly introduced macroprudential policies in
Korea have seen some initial success in moderating the sensitivity of capital ows to external
factors.
5 Robustness Checks
5.1 Evidence from Other Countries in Asia.
Evidence of change in incremental sensitivity cannot be taken as conclusive proof of the eect
of the macroprudential policy, since the experiment is simply to look before and after June2010. However, we complement our incremental sensitivity analysis for Korea by comparing
the results for Korea with those of other countries in the region, as well as examining evidence
for any structural breaks in the way that our variables of interest impact capital ows, so that
we may place the results on Korea into context. Through this exercise, we may ascertain the
extent to which the results for Korea are shared by other Asian countries, and hence give clues
on any regional variations in our sample.
We therefore run panel regressions analogous to those in Table 2 but for each of the following
countries: Australia (AUS), Indonesia (IDN), Malaysia (MYS), Philippines (PHL), Thailand
(THA) and Vietnam (VNM).
The choice of Australia follows from the similarity of Australias banking sector in its open-
ness and its reliance on wholesale funding. Australia also has a similar sized economy to Korea,
a fellow member of the G20 grouping of countries and the Australian dollar has some of the same
attributes as the Korean won as a barometer of the risk-taking behavior in global capital mar-
kets. The other countries (Indonesia, Malaysia, Philippines, Thailand and Vietnam) are chosen
as they are collectively known as the Big 5 ASEAN countries and share some similarities to
Korea in terms of openness to similar regional forces.
We focus on the Interoce variable as the main global factor capturing the funding ac-
26
-
7/28/2019 CL2 HynSongShin Korea Macropru
27/36
Table 7. Sensitivity of bank capital ows to other Asian economies. This table reports panel regressionsfor capital ows designed to identify incremental sensitivity of a particular country. Separate specications are
run for Australia (AUS), Indonesia (IDN), Malaysia (MYS), Philippines (PHL), Thailand (THA) and Vietnam(VNM). In each case, the dependent variable is bank capital ows as measured by the quarterly log dierenceof external loans (BIS Table 7A). Expanatory variables include the global and local variables in Table 2,interactions of Interoce with the time dummy for the period from June 2010 and with the dummy for theparticular country examined. Standard errors are clustered at the country level and are reported in parantheses.
Coe. 1 2 3 4 5 61 Interoce 0.0122** 0.0112** 0.0108** 0.0117** 0.0118** 0.0109*
[0.0053] [0.0053] [0.0053] [0.0053] [0.0054] [0.00532 Interoce*Post 2010 0.0526** 0.0539** 0.0529** 0.0502** 0.0514** 0.0534*
[0.0220] [0.0222] [0.0220] [0.0222] [0.0223] [0.02233 Interoce*AUS -0.0509***
[0.0053]4 Interoce*AUS*Post 2010 0.0860***
[0.0159]
AUS*Post 2010 0.0410***[0.0099]
3 Interoce*IDN -0.0091*[0.0053]
4 Interoce*IDN*Post 2010 0.0228[0.0162]
IDN*Post 2010 0.0737***[0.0086]
3 Interoce*MYS 0.0083*[0.0049]
4 Interoce*MYS*Post 2010 0.0696***[0.0157]
MYS*Post 2010 0.0687***[0.0090]
3 Interoce*THA -0.0314***
[0.0053]4 Interoce*THA*Post 2010 0.1965***[0.0162]
THA*Post 2010 0.1665***[0.0084]
3 Interoce*PHL -0.0333***[0.0047]
4 interoce*PHL*Post 2010 0.1394***[0.0160]
PHL*Post 2010 0.0931***[0.0085]
3 Interoce*VNM 0.0165*[0.0052
4 Interoce*VNM*Post 2010 0.0464*[0.0165
VNM*Post 2010 0.0342*[0.0088Controls Y Y Y Y Y Y
27
-
7/28/2019 CL2 HynSongShin Korea Macropru
28/36
Table 8. F tests for coecient restrictions in Table 7. This table presents F tests for coecient restrictionsin Table 7 for the interaction of Interoce with the time dummy for the period from June 2010 and with thedummy for the particular country examined. We examine Australia (AUS), Indonesia (IDN), Malaysia (MYS),Philippines (PHL), Thailand (THA) and Vietnam (VNM). The null hypothesis (2) + (4) = 0 corresponds tothe statement that the sensitivity of capital ows to that country with respect to Interoce does not changeafter June 2010. The null hypothesis (3) + (4) = 0 corresponds to the statement that there is no dierence insensitivity to Interoce between the country and the other countries in the 48 country sample after June 2010.p-values are reported for each null hypothesis.
(1) (2) (3) (4) (5) (6)
Null hypothesis Country
AUS IDN MYS THA PHL VNM
(2) + (4) = 0 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
(3) + (4) = 0 0.0278 0.3798 0.0000 0.0000 0.0000 0.0003
(1) + (2) = 0 0.0057 0.0055 0.0060 0.0084 0.0071 0.0063
(1) + (3) = 0 0.0000 0.3711 0.0000 0.0000 0.0000 0.0000(1)+(2)+(3)+(4)=0 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
tivities of global banks, and follow the empirical procedure reported in Table 2 using the
country dummy and Post 2010 dummy, as well as the interaction between the two. The re-
sults are reported in Table 7. For economy of presentation, we report only the key interac-
tion coecients. For each country, we report the coecient on Interoce*Country and on
Interoce*Country*Post 2010. We assign the same numbering to the interaction dummies
coecients as we did for Table 2 and Table 3.The evidence on Interoce is striking. In contrast to Korea, all countries in Table 7 except
Indonesia show positive coecients on the triple interaction term Interoce*Country*Post
2010, meaning that Australia, Malaysia, Thailand, Philippines and Vietnam became more sen-
sitive to uctuations in global bank funding conditions after June 2010. Even in the case of
Indonesia, the coecient is insignicant, rather than being negatively signicant as in Korea.
We conduct F-tests for structural change, and the results are reported in Table 8. The
numbering convention for the null hypotheses remains the same as before. The null hypothesis
(2) + (4) = 0 corresponds to the statement that the sensitivity of the country to Interoce
does not change after June 2010. We see that this hypothesis is resoundingly rejected for all
28
-
7/28/2019 CL2 HynSongShin Korea Macropru
29/36
four local variables. However, unlike in the case of Korea, the null hypothesis (2) + (4) = 0
is rejected for the opposite reason from Korea. In these other Asian countries, capital ows
became more sensitive to global factors after 2010, rather than less sensitive.
The null hypothesis (3) + (4) = 0 corresponds to the statement that there is no dierence
between the country and the other countries in the full sample after June 2010. We see that the
null hypothesis (3) + (4) = 0 is strongly rejected for Australia, Malaysia, Thailand, Philippines
and Vietnam. It is only for Indonesia that the null (3) + (4) = 0 cannot be rejected.
Taken together, the evidence suggests that there has been a markedly lower sensitivity to
global funding on capital ows in Korea, even when all other comparable countries in the region
saw the opposite eect - toward an increased sensitivity to global factors in the determination
of capital ows. Our results are suggestive that the introduction of macroprudential policies inKorea served to mitigate the notorious sensitivity of Korea to external nancial conditions.
5.2 Dating the Structural Change
So far, we have relied on the a priori knowledge of policy announcements and press reports to
choose the threshold date for the macroprudential policies to June 2010. We will now adopt a
more agnostic procedure in dating the structural break by examining at each date the evidence
for structural break at that date, and then compiling the evidence for the whole period after
the start of 2010.
Specically, we run the following panel specication.7 Starting from the benchmark panel
regression, we introduce the interaction dummies:
Interoce*Korea*Pre, Interoce*Pre, (5)
where Pre is the dummy variable that takes the value 1 if the date is before the end of 2009,
and zero otherwise. We also introduce the interaction dummies
Interoce*Korea*Q(n), Interoce*Q(n), (6)
7 We are grateful to a referee for suggesting this specication.
29
-
7/28/2019 CL2 HynSongShin Korea Macropru
30/36
Table 9. Dating the structural break. This table reports panel regressions with capital ows as thedependent variable. Explanatory variables include global and local variables in Table 2, and interaction dummiesInteroce*Korea*Pre and Interoce*Pre, where Pre = 1 up to the end of 2009, and zero afterwards. Alsoincluded are Interoce*Korea*Q(n) and Interoce*Q(n) where Q (n) = 1 in the nth quarter from 2010:Q1and zero otherwise. We run separate panel regressions for each n and report their coecients below.
Coecient std err t date
a0 interoce*Korea*Pre 0.036456 0.005497 6.63 Pre Mar-10a1 interoce*Korea*Q1 -1.99128 0.302513 -6.58 Mar-10a2 interoce*Korea*Q2 -0.38397 0.078521 -4.89 Jun-10a3 interoce*Korea*Q3 -1.50035 0.20109 -7.46 Sep-10a4 interoce*Korea*Q4 -1.2222 0.288706 -4.23 Dec-10a5 interoce*Korea*Q5 0.64302 0.15142 4.25 Mar-11a6 interoce*Korea*Q6 -0.08409 0.014951 -5.62 Jun-11a7 interoce*Korea*Q7 -0.03439 0.025341 -1.36 Sep-11a8 interoce*Korea*Q8 -0.04917 0.020808 -2.36 Dec-11a9 interoce*Korea*Q9 0.082683 0.074849 1.1 Mar-12b0 interoce*Pre 0.013826 0.005207 2.66 Pre Mar-10b1 interoce*Q1 1.417899 0.300799 4.71 Mar-10b2 interoce*Q2 0.348851 0.080573 4.33 Jun-10b3 interoce*Q3 1.148828 0.194686 5.9 Sep-10b4 interoce*Q4 -0.69569 0.291348 -2.39 Dec-10b5 interoce*Q5 -0.02685 0.147972 -0.18 Mar-11b6 interoce*Q6 0.031209 0.014856 2.1 Jun-11b7 interoce*Q7 0.069443 0.026399 2.63 Sep-11
b8 interoce*Q8 0.071907 0.019245 3.74 Dec-11b9 interoce*Q9 0.064104 0.06774 0.95 Mar-12
for quarter n starting from the rst quarter of 2010, and where Q (n) is the dummy that takes
the value 1 in quarter n and zero otherwise. We then run the benchmark panel regression with
the inclusion of the two triple interaction terms in (5) and (6), and report their coecients in
Table 9.
By examining the coecients and the standard errors for these variables for each quarter
beginning in 2010, the rationale of this procedure is to obtain a more precise dating of thestructural break for Korea, and to examine how uniform the lower sensitivity to global factors
30
-
7/28/2019 CL2 HynSongShin Korea Macropru
31/36
are over the post-2010 period.
From the t-statistics reported in Table 9, we see that a0 > 0, and so the sensitivity of Korea
to Interoce prior to 2010 was strongly positive. However, for the coecients a1 to a9, we see
that all but one is negative and signicant. The only exception is March 2011. We observe the
reverse pattern for the t coecients, which are all positive and signicant with the excpetion
of4:
From this evidence, we can conclude that Korea became less sensitive compared to global
factors compared to the other countries in our sample of 48 countries. The evidence is consis-
tently present throughout the period after 2009. These results give some cause for reassurance
that our earlier results that rely on the partitioning of the sample period into before and after
subsamples does not rely on outlier observations to mask the overall eect. The evidence pointsto the impact of the macroprudential policies falling uniformly over the whole period from 2010.
6 Conclusions and Directions for Further Research
Given the accumulated evidence assembled in this paper, the following conclusions seem justied
concerning the impact of Koreas macroprudential policies.
First, there is evidence from a variety of approaches that the sensitivity of capital ows to
global factors reduced substantially in the case of Korea from 2010. We have seen this boththrough the signs of the interaction dummies, through F-tests of structural breaks, as well as
the co-movement measurements using our Beta(i; t) variables.
Remarkably, the dampening of sensitivity in Korea sits side-by-side with evidence that Ko-
reas experience is dierent from group of comparable Asian countries. For these Asian coun-
tries, they saw an increase in the sensitivity of their capital ows to global factors after 2010.
Since the original ndings for Korea were relative to the global comparison group, the increased
sensitivity of Australia, and other Asian comparison countries is representative of global trends
as a whole. Given such a backdrop, the conspicuous drop in the sensitivity of capital ows into
Korea is all the more notable.
31
-
7/28/2019 CL2 HynSongShin Korea Macropru
32/36
Finally, we have seen that the lower sensitivity of Koreas capital ows to global factors after
2010 is an eect that is uniformly present throughout the post-2010 period, rather than relying
on one or two outliers to get the overall sign.
The results in our paper reiterate a number of broader lessons. The evidence in our paper
suggests that the driving force behind banking sector capital ows is the leverage cycle of the
banking sector, through the interaction of the supply and demand of wholesale bank funding.
Our ndings reinforce the argument in Borio and Disyatat (2011), Obstfeld (2012a, 2012b) and
Gourinchas and Obstfeld (2012) on the importance of gross capital ows between countries in
determining nancial conditions, especially the gross ows intermediated by the banking sector.
Bank capital ows have also been pivotal in the European nancial crisis. The credit boom
in countries such as Ireland and Spain were nanced primarily by capital ows through thebanking sector (see Allen, Beck, Carletti, Lane, Schoenmaker and Wagner (2011) and Lane and
Pels (2011)). Therefore, the mechanisms outlined here on the link between capital ows and
leverage are relevant in understanding the European crisis, also.
Our ndings highlight the role of nancial intermediaries in driving uctuations in risk
premiums and nancial conditions, especially in connection with the growing use of wholesale
bank funding.
The procyclicality of banking sector capital ows poses challenges in setting policy and
regulatory responses. The cross-border spillovers associated with banking sector ows highlights
the importance of international coordination in banking regulation and in monetary policy, but
such coordination is not straightforward to design or implement, even when the interests of the
relevant countries are congruent. Moreover, even when coordination is globally optimal, it still
may generate tensions with national governance.
In the absence of eective international coordination, a second best approach (that takes
the spillovers as given) would be appropriate in designing a framework to mitigate the risks of
cross-border ows at the national level. The recent report by the Committee on International
Economic Policy Reform (CIEPR (2012)) describes the considerations that are relevant in setting
policy on capital ows in a second best world. The macroprudential policies introduced in Korea
32
-
7/28/2019 CL2 HynSongShin Korea Macropru
33/36
in 2010 can be viewed in this context. The Korean measures should also be seen in the context
of the broader debate on the design of policies toward nancial stability. Although the term
macroprudential is now commonly encountered in the policy world, empirical studies on their
eectiveness have been comparatively less common. In this context, our empirical results may
be a useful input in more detailed studies of the tradeos involved in macroprudential policy
design.
33
-
7/28/2019 CL2 HynSongShin Korea Macropru
34/36
References
Adrian, Tobias and Hyun Song Shin (2010) Liquidity and Leverage, Journal of FinancialIntermediation, 19, 418-437
Adrian, Tobias and Hyun Song Shin (2012) Procyclical Leverage and Value-at-Risk FederalReserve Bank of New York Sta Report 338,http://www.newyorkfed.org/research/sta_reports/sr338.html
Baba, Naohiko, Robert N. McCauley and Srichander Ramaswamy (2009) US Dollar MoneyMarket Funds and Non-US Banks BIS Quarterly Review, March 2009, 65-81.
Bank for International Settlements (2010) Funding patterns and liquidity management of in-ternationally active banks CGFS Papers 39, Committee on the Global Financial System,http://www.bis.org/publ/cgfs39.pdf
Bank for International Settlements (2011) Global liquidity concept, measurement and policyimplications CGFS Papers 45, Committee on the Global Financial Systemhttp://www.bis.org/publ/cgfs45.pdf
Borio, Claudio and Piti Disyatat (2011) Global imbalances and the nancial crisis: Link or nolink? BIS Working Papers No 346 http://www.bis.org/publ/work346.pdf
Brunnermeier, Markus and Lasse Pedersen (2009) Market Liquidity and Funding Liquidity,Review of Financial Studies, 22, 2201-2238.
Brunnermeier, Markus and Yuliy Sannikov (2011) The I-Theory of Money working paper,
Princeton University
Bruno, Valentina and Hyun Song Shin (2011) Capital Flows, Cross-Border Banking and GlobalLiquidity working paperhttp://www.princeton.edu/~hsshin/www/capital_ows_global_liquidity.pdf
Calvo, Guillermo A., Leonardo Leiderman, and Carmen Reinhart, Capital Flows to DevelopingCountries in the 1990s: Causes and Eects, Journal of Economic Perspectives, 10, Spring 1996,123-139.
Cetorelli, Nicola and Linda S. Goldberg (2009) Banking Globalization and Monetary Trans-
mission Forthcoming in Journal of Finance.
Cetorelli, Nicola and Linda S. Goldberg (2010) Global Banks and International Shock Trans-mission: Evidence from the Crisis, Federal Reserve Bank of New York Sta Report #446.
34
-
7/28/2019 CL2 HynSongShin Korea Macropru
35/36
Chung, Kyuil, Hail Park and Hyun Song Shin (2012) Mitigating Systemic Spillovers fromCurrency Hedging National Institute Economic Review, 221(1) R44-R56, National Institute ofEconomic and Social Research, London.
Claessens, Stijn and Swati R. Ghosh (2012) Macro-Prudential Policies: Lessons for and fromEmerging Markets forthcoming in proceedings of KDI/EWC conference Financial Regulationson International Capital Flows and Exchange Rates, July 2012,http://www.g24.org/TGM/Macro-Prudential%20Policies.pdf
Claessens Stijn, Neeltje Van Horen, Tugba Gurcanlar and Joaquin Mercado Sapiain (2008)Foreign Bank Presence in Developing Countries 1995-2006: Data and Trends working paper,http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1107295
Committee on International Economic Policy Reform (2012) Banks and Cross-Border CapitalFlows: Policy Challenges and Regulatory Responses, Brookings Institution, September 2012.
Danielsson, Jon, Hyun Song Shin and Jean-Pierre Zigrand (2012) Procyclical Leverage andEndogenous Risk working paper
Forbes, Kristin J. and Francis E. Warnock (2011) Capital Flow Waves: Surges, Stops, Flightand Retrenchment working paper, MIT Sloan School and University of Virginia.
Galati, Gabriele and Richhild Moessner (2012) Macroprudential policya literature reviewJournal of Economic Surveys, Wiley Blackwell.
Gourinchas, Pierre-Olivier and Maurice Obstfeld (2012) "Stories of the Twentieth Century for
the Twenty-First,"American Economic Journal: Macroeconomics
, 4(1), 226-65.Hahm, Joon-Ho, Hyun Song Shin and Kwanho Shin (2011) Non-Core Bank Liabilities andFinancial Vulnerability forthcoming in Journal of Money, Credit and Banking
International Monetary Fund (2012) The Interaction of Monetary and Macroprudential Poli-cies: Background Paper IMF Background paper,http://www.imf.org/external/np/pp/eng/2013/012713.pdf
Lane, Philip and Gian Maria Milesi-Ferretti (2007). "The External Wealth of Nations MarkII: Revised and Extended Estimates of Foreign Assets and Liabilities, 19702004" Journal ofInternational Economics, 73, 223-250.
Lane, Philip and Barbara Pels (2011) "Current Account Balances in Europe." working paper,Trinity College Dublin
35
-
7/28/2019 CL2 HynSongShin Korea Macropru
36/36
McGuire, Patrick and Goetz von Peter (2009) The US Dollar Shortage in Global Banking BISQuarterly Review, March 2009, Bank for International Settlements.
Obstfeld, Maurice (2012a) Does the Current Account Still Matter?, 2012 AEA Ely Lecture,American Economic Review, 102(3), 1-23.
Obstfeld, Maurice (2012b) Financial Flows, Financial Crises, and Global Imbalances, Journalof International Money and Finance, 31, 469480
Pradhan, Mahmood, Ravi Balakrishnan, Reza Baqir, Georey Heenan, Sylwia Nowak, CeydaOner, and Sanjaya Panth (2011) Policy Responses to Capital Flows in Emerging Markets IMFSta Discussion Note, 11/10, International Monetary Fund
Reinhart, Carmen M and Kenneth S Rogo (2009), This Time is Dierent: Eight Centuries ofFinancial Folly, Princeton, NJ: Princeton University Press
Schularick, Moritz and Alan M. Taylor (2012) Credit Booms Gone Bust: Monetary Policy,Leverage Cycles, and Financial Crises, 1870-2008. American Economic Review 102 (2012),forthcoming
Shin, Hyun Song (2010) Non-Core Liabilities Tax as a Tool of Prudential Regulation policymemo, http://www.princeton.edu/~hsshin/www/NonCoreLiabilitiesTax.pdf
Shin, Hyun Song (2012a) Adapting Macroprudential Policies to Global Liquidity ConditionsCentral Bank of Chile working paper 671,http://www.bcentral.cl/eng/studies/working-papers/pdf/dtbc671.pdf
Shin, Hyun Song (2012b) Global Banking Glut and Loan Risk Premium IMF EconomicReview, 60: 155-192
Shin, Hyun Song and Kwanho Shin (2010) Procyclicality and Monetary Aggregates, NBERdiscussion paper w16836, http://www.nber.org/papers/w16836.
Xiong, Wei (2001) Convergence Trading with Wealth Eects: An Amplication Mechanism inFinancial Markets, Journal of Financial Economics 62, 247-292
36